The invention relates generally to semiconductor integrated circuits and more specifically to management of performance parameters of integrated circuits.
Ideally, semiconductor manufacturing processes would produce perfectly identical semiconductor devices. However, impurities and process variations result in semiconductors devices, even those from the same wafer, having different characteristics. An example of the different characteristics include differences in timing characteristics. Variations in timing characteristics result in different operating speeds for the components. Integrated circuit producers typically grade integrated circuits in different speed categories, or xe2x80x9cbins,xe2x80x9d and sell them as different components at different prices. The high-speed parts are typically sold at a higher price. Computer manufacturers buy different xe2x80x9cbinxe2x80x9d CPUs at different prices and offer them in models with different price points, e.g., PCs based on PENTIUM III 550 MHz CPUs are sold at a higher price than the ones based on PENTIUM III 500 MHz CPUs.
To reduce the manufacturing and inventory issues, computer manufacturers use the same mother boards for end products based on these different speed CPUs. A clock generator on the motherboard generates a fixed frequency clock that is fed to the CPU. The CPU has an internal register that determines a multiplier ratio between the external clock frequency and the CPU core frequency. Computer manufacturers, at the time of system integration, program the system with the correct multiplier to run the CPU core at its rated speed.
This approach has several problems. Firstly, the additional step involved in programming the system increases the manufacturing cost. Secondly, rouge computer manufacturers can and do over-clock systems and sell computers based on lower speed xe2x80x9cbinxe2x80x9d CPUs as if they were based on higher speed xe2x80x9cbinxe2x80x9d CPUs. This compromises the system integrity, as well as harming CPU manufacturers. Thirdly, gray market peddlers re-stamp parts as high speed parts. Thus, integrated circuits, as currently configured, suffer from many disadvantages.
Another shortcoming of integrated circuits and the systems in which they are used is that they are typically configured to operate at fixed performance and power dissipation levels and cannot be dynamically altered. These fixed performance and power dissipation levels can result in undesirable effects on other operating parameters. Such integrated circuits and systems are incapable of dynamically adapting the performance and power dissipation levels to be more suitable for the manner in which users are using the integrated circuits and systems.
Another problem that arises as more processing capability is built into integrated circuits is that the power dissipated by the integrated circuits increases. Traditional techniques for allowing integrated circuits to operate with increased power dissipation include the use of heat sinks, fans, and thermoelectric cooling devices to cool the integrated circuits. Heat sinks are bulky and add to the cost of the systems. Fans decrease reliability by introducing a moving part and increase audible noise from the system, as well as increasing cost and complexity and reducing battery life for battery-operated systems. Thermoelectric cooling devices are expensive, consume large amounts of power, generate additional heat, and reduce battery life for battery-operated systems. Thus, traditional thermal management techniques introduce additional complications and costs and are constrained in their effectiveness.
In summary, traditional techniques for producing and operating integrated circuits and systems in which they are incorporated have substantial drawbacks. Moreover, such drawbacks tend to be exacerbated by the manner in which modern integrated circuits and systems are used, for example, in battery-operated systems and in general-purpose computing systems capable of performing a great number of widely varied applications.